🚀 go-pugleaf

> |> There's an addendum to the above-mentioned article, in both December '95
> |> and February '96 issue of Byte.  They discuss a bug (more of a
> |> non-optimal malloc library routine) which caused poor ByteMark
> |> performance to be given for the Intel machines.  There are revised
> |> figures available in the latter article showing better performance for
> |> the P6.
> 
> Nice try.  Here are the corrected figures:
> 
> Let's talk actual systems, not estimated processor benchmarks.
> January 1996 Byte, Page 140 Graph "Raw Performance: The VIS-less Ultra
> I".  Sorted in order of integer performance:
> ----------------------------------------
> DeskStation Raptor 3 266 MHz Aplha 21164:                3.75, 4.49
> IBM Personal Computer Power Series 850, 133 MHz PPC 604: 2.90, 2.92
> Sun Ultra I Creator 3D Model 170 167 MHz UltraSparc:     2.18, 1.54
> Intel Reference System, 150 MHz Pentium Pro:             2.10, 2.60
> Netpower Fastseries MP 200 MHz Mips R4400:               2.03, 2.08
> 
> PPC 604 is faster than Pentium Pro, even though the PPro is running at
> a higher clock speed.
> 
> |> I think it's important to know which compiler (with which optimizations
> |> turned on) and which system (core logic, cache, memory, bus) are used
> |> before one can truly judge performance/benchmark comparisons. IMHO.
> 
> Agreed.
> 

I apologize for the misunderstanding, what I said wasn't very clearly
stated.  The article concerning the BYTEMark bug is at:

http://www.byte.com/art/9512/sec4/art2.htm

I pointed out this reference since the November 95 article seems to be
passed around as gospel, quite often without mention of the bug.
Comparative differences between the 133MHz PPC and the 150MHz P6 isn't
as stark with the corrected BYTEMark benchmarks.

The article which I referenced concerning the better P6 performance is
in the February issue of BYTE. It's not on the net, so I can't provide a
HTML reference; the article does give computed (not estimated) values.
The faster performance there was for a 200MHz (and not 150MHz) version
of the P6.  Note that that's a faster clock speed than the 133MHz PPC,
but how that's an issue isn't clear, since Intel seems to have the
superior fab processing to ramp up the clock to faster  levels to keep
up performance wise.  What about the faster PPC coming out?  Well, then
they'll be a faster P6 then an even faster PPC, then a even faster P6,
and so on and so on through the 620 and the MMX and P7...  

I didn't intend my previous post to add more wood to the PPC vs. P6
flame war.  They're both very nice processors with distinct advantages
for different applications, with performance pretty comparable in the
specific application of running certain benchmarks.

I did want to point out that I (IMHO) disagree with people posting
benchmark references without being clear about the systems (cpu, bus,
cache, memory, core logic, etc) and the compilers (optimization,
library, etc.) which are used. For instance, one interesting page to
look at is

http://www.mot.com/SPS/PowerPC/library/technical_papers/performance/mac_bm/b12.html

which show PowerMAC 9500/132 BYTEMark floating-point results using
different compilers (they do clearly indicate the system and compiler
optimizations used).  The diffences in performances achieved there are
quite stark, in fact more so than the comparative differences between
the various systems cited in the previous posting.  Caveat emptor.

Peace and good will,

Tetsuji Ueda

----------------------------
My other computer is a Cray!
----------------------------